Investigators cite observer variability as a problem in using crackles to diagnose asbestosis. We measured agreement on the presence or absence of crackles noted during auscultation of 64 asbestos-exposed workers in order to clarify this question. There was 89 percent agreement between two observers who simultaneously examined subjects breathing from functional residual capacity (FRC). Kappa (kappa), a statistic accounting for chance agreement, was 0.73. Unanimous agreement between four observers who listened to tape recordings of the breath sounds was 81 percent (kappa = 0.69). When the subjects breathed from residual volume (RV) there was 78 percent (kappa = 0.53) and 67 percent (kappa = 0.60) agreement, respectively. Comparing direct to tape-playback auscultation, there was 90 percent (kappa = 0.77) and 84 percent (kappa = 0.58) intraobserver agreement when the subjects breathed from FRC and 90 percent (kappa = 0.79) and 75 percent (kappa = 0.39) when they breathed from RV. We conclude that observer variability is sufficiently low to allow trained observers to monitor asbestos-exposed workers for crackles directly and during tape-playback auscultation.